[Pacemaker] resource go in unmanaged mode when node startup

[Andreas Kurz]

Am 02.09.11 18:47, schrieb Umberto Carrara:
> hi,
> I have a cluster with kvm vms and libvirt+iscsi, everything is fine except 
> when i restart a node, when corosync service start, the cluster try move 
> running vms resource in alive node before the resources, iscsi and libvirt, 
> are ready 
> 
> this rules
> 
> colocation domainonservice inf: file serviceclone
> order servicebeforedomain inf: serviceclone:start  file
> 
> within
> 
> resource-stickiness="1000"
> 
> should prevent this, right?
> 
> 
> the only thing to do is a cleanup of resources.
> 
> is a problem of timeout or wrong config?

this is a chicken-and-egg problem. libvirtd has to be started for
VirtualDomain RA probing to work.

If you start libvirtd as a cluster resource this is too late, start it
via init. On probing errors Pacemaker can't reliable tell if a resource
is started outside cluster control, so for safety reasons a restart is
triggerd -- stop on all nodes, rstart it on only one node:

Sep  2 18:14:05 host2 pengine: [1735]: ERROR: native_create_actions:
Resource file (ocf::VirtualDomain) is active on 2 nodes attempting recovery
Sep  2 18:14:05 host2 pengine: [1735]: WARN: See
http://clusterlabs.org/wiki/FAQ#Resource_is_Too_Active for more information.

Regards,
Andreas

> 
> in attach my crm config and log
> 
> Thanks to all in advance
> 
> Umberto
> 
> 
> 
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker